How many publications to display in the UI, what information to provide (options to select abstract, PMID, title, authors, full pub?)

[karafecho]

The QotM surfaced a (known) issue related to publication support. Specifically, how should Translator balance the variance among users, in terms of the amount (just a few papers, as many as you have) and type (PMID, PMID + abstract, PMID + full publication) of publication support they want to see? Specificity of publications that are provided? How do users apply publication support? Confidence in edge? In depth exploration? Both, depending on where the publication support is originating?

[sierra-moxon]

from TAQA:

• some F&F users say there are too many papers (might need to read or skim all the papers to understand the statement).
• not something UI can decide; hard to choose just one publication to display.
• maybe it's a user education issue?
• are the papers sorted? - not yet, the date would be one to sort on, but its provided in too many formats to be able to sort.

[sierra-moxon]

For UCWG: What does a user want when they see a "publication" or a large number of publications for a statement? UI is hesitant to exclude publications, but would like direction on how to filter large number of publications on a statement that would be most helpful for a user.

[karafecho]

Related, there is a a bias toward recent publications on 'hot' topics or heavily researched topics, e.g., COVID, cancer and a desire by users to filter by type of publication (e.g., case study, review). Filter options would help to address these issues.

[sandrine-m]

I'll bring the issue to the UCWG at next meeting. Thank you!

[sierra-moxon]

From UCWG:

• can we categorize the pubs by the diversity of publication? (maybe from many different journals - rank these higher) ?
• based on author lists, knowledge should be reproduced from many different labs to be considered "better evidence" ?
• could use MeSH terms to compare pub content automatically -- "more similar" = "more better"
• the resulting graph might be something you filter on - one hop graph is not as interesting? maybe a 3-4 hop graph might give you more priority and move that evidence to the top of the list.
• the main issue is having a clear way to filter the evidence just like other components (maybe a slider bar). Only publications post 2019, only publications from these journals, etc.

Action items:

• do we have all the attributes we need to do this filtering and ranking?
• is there code already to compare attributes between publications for an edge, so that we can rank the pubs (or show the top 10, etc)
• ability to filter/cluster on terms or attributes (like date of publication, journal) is a UI issue.
  • PMI users have also asked for some filtering terms like the "snippet" and the "date" - so those are there (in the UI) already.

discussion on action items:

• text mining server shows publication attributes (maybe not MeSH terms, but some abstract details, authors etc.)?
• TMKP might already have a measure for us -- what are the terms that have already been tagged in the abstract, and can we use the similarity of this list from one pub to another to compare.?

[sierra-moxon]

@bill-baumgartner - some discussion about this particular ticket at the UCWG (summarized above). It sounds like TMKP already has a data source that stores publication attributes consistently and which is being used already in the UI.

Does this data source also have MeSH terms? Does it already have any calculations or comparison of attributes between publications? Can we use the data in the data source to write some code that would compare x number of publications from a single edge, for breadth and depth of knowledge (calculated by similarity of MeSH terms, author lists etc)? This may be too hard to answer without context :) -- we could discuss at TAQA if needed.

[tursynay]

UCWG does not have any action items for this issue, however to be discussed at the relay

[bill-baumgartner]

@sierra-moxon apologies for not seeing this earlier. We don't currently provide MeSH terms in our publication metadata API, but that would be an easy addition. We also don't supply any comparison calculations between publications but would be willing to modify the API based on the needs to the UI/Consortium.

Current metadata provided:

• document id
• year published
• month published
• day published
• journal
• journal abbreviation
• volume
• issue
• article title
• article abstract

[sierra-moxon]

Thank you so much @bill-baumgartner @sandrine-m and Paul C and the UCWG for the ideas here. I've put this into the "TACT" meeting bucket so that we can get an executive decision on whether the strategy for "fixing" this issue is reasonable. @tursynay

[sstemann]

@sierra-moxon can this one be closed? the UI is leveraging the TMKP service. I'm not sure what the TACT team or strategy is for the other parts but i'm not sure it's still a "UI - display" issue.

[sierra-moxon]

closing.